Its been a mystery and often have people asked themselves how the f**k we come up with the match ratings the way we do.
I'm happy that i can now put a large portion of that to an end - we are finally confident enough with the system that we will make it public.
The Basics:
The ratings depend to 95% on the tv viewerpeak that everyone can see below the server listings on the matchpage. We constantly calculate a scale that puts matches with most viewers into the 3 star area and adjusting everything else according to that.
The Math:
1. We find the match with the most viewers in the past 3 Month - thats currently an allstars match with a viewerpeak of 906
2. We take the square root of that number to get those top scorers a little closer to reality ;-) - thats about 30 atm
3. We devide this area into 10 ratings - 1 is 0-3, 2 is 3-6, ... 10 is 27 to 30
Thats it basically - if we have a match, we take the square root of its viewerpeak and check into which area it falls, giving it a rating from 1 to 10 (internally)
This rating is then translated into stars in the following way:
1: 0 Stars
2-3: 1 Star
4-6: 2 Stars
7-10: 3 Stars
Some numbers:
Again based on current figures (for et only) (and they change constantly) this is how it decides
0 - 20 viewers: 0 Stars
20 - 110 viewers: 1 Star
110 - 380 viewers: 2 Stars
380 - 900+ viewers: 3 Stars
So you got a time machine?
No we don't - although we'd like to simply look into the future and check how the viewerpeak will turn out we can not :-(
Viewer peaks for matches in the future are estimated from statistical data gained from previous matches of the same teams or in the same league. In most cases this is pretty accurate but as you have probably noticed already gtv sometimes adjusts the number of stars up or down during or after a match (when viewerpeak is known) - especially for borderline cases.
So next time you see "lol X Stars" in the comments, just link them here ;-)
arni
Forum » About GamesTV » GamesTV Coverage » Match ratings explained
 arni
|
Match ratings explained on
28.10.09 21:47
|
 skooli
|
i almost fell asleep reading this
|
 msh100
|
Seems like a realistic method
|
|
arni
|
thats because its not interesting at all - but some people seem to wanna know :p |
4drenalin3
|
It seems quite accurate indeed, but is there any specific reason why you take the match with the most viewers of the past 3 months and not the past week, for example?
|
|
arni
|
yes, we believe that thats about as long as people will think back knowing that that was a real top match - everything else should be adjusted to that |
ng
|
will do so ;-) |
 eshan
|
nice!
|
 phase
|
no more stars for my clan at gtv ;(
|
 caybro
|
wait, we will play against you :) |
cl4ym4n
|
coz youre too noob, kapusta ;D |
|
Rhand.
|
380 viewers seems pretty low for 3 stars, or has ETTV really gone downhill so much that you can actually speak about a topmatch when 380 viewers watch?
|
 cocaiNe+
|
hey nice, i am just curious
"The ratings depend to 95% on the tv viewerpeak" whats the other 5%??? |
andyF1
|
Luck !!! or more likely the league that the match / event is part of. |
w0nd3r
deleted this post of
reVEEe
(14.12.09 20:05)
|
 domrod93
|
I guess so. Look : http://www.gamestv.org/event/14850-ec-winners-vs-tag/ => Viewer Peak: 333 http://www.gamestv.org/event/14389-team-yoyotech-vs-ec-winners/ => Viewer Peak: 253 http://www.gamestv.org/event/14623-ec-winners-vs-masculine-mans/ => Viewer Peak: 442 http://www.gamestv.org/event/14734-tag-vs-masculine-mans/ => Viewer Peak: 287 http://www.gamestv.org/event/14754-sleeperz-vs-tag/ => Viewer Peak: 325 So the chosen value seems pretty realistic compared to audience. By the way, nice job Arni :). |
|
arni
|
the 3 star value is lower now btw - it hink its around 280
|
|
essAh
|
Site Sups should be able to change the match rating manually, would be a nice feature.
|
|
skooli
|
ratings shouldnt be set manually at all.
|
|
ng
|
nope |
 |
This Thread is closed and cannot be commented on anymore. |
